JP2003109328A - Storage device and its method for correcting error - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a means for estimating an error position by utilizing NRZ data itself in order to avoid omission and rising of a TA flag and an error flag caused by difficulty of optimally setting a slice. SOLUTION: The same sector on a magnetic disk is read several times, an obtained plurality of pieces of NRZ data are stored, the pieces of NRZ data are compared in a byte unit, it is decided that an error occurs at a byte position where reproduced NRZ data is different in each reading operation, and the position is utilized as an erasure pointer for erasure correction.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an error correction method performed when a read error occurs in a magnetic disk device.

[0002]

2. Description of the Related Art The recording density of a magnetic disk device is increasing year by year, and accordingly, an error generated when a defect of the same size exists on a medium is becoming visible as a large error length. . When a large error occurs, a simple means to detect and correct the error is to increase the error correction code added to the data. However, this method lowers the format efficiency, which leads to a vicious circle in which the recording density must be further increased.

In other words, as an error correction code (2 × n)
When data is written with the Reed-Solomon code of bytes added, n bytes are used to identify the error position.
The maximum error correction capability is n bytes. Therefore (4 ×
If data is written by adding a (n) byte Reed-Solomon code, it is possible to correct an error of up to (2 × n) bytes. However, since the capacity for recording the increased (2 × n) bytes of data decreases, increasing the recording density to compensate for the decrease increases the occurrence of errors and increases the correction capability. The merit is reduced.

There is a method called erasure correction as a method for improving the correction capability without changing the length of the error correction code to be added.

FIG. 2 shows a system configuration when erasure correction is applied. The part surrounded by the broken line is generally HDC
Provided as a function of.

The NRZ data reproduced by the R / W channel 201 is usually temporarily stored in the buffer memory 202. From the NRZ data extracted from the buffer memory 202 by the HDC, the syndrome generator 203 generates a syndrome. The decoder 204 uses the erasure position polynomial (eras
ure-locator polynomial) is generated.

A TA flag and an error flag issued by the R / W channel are used to specify the error position.

The TA flag is a thermal asperity (=
This is a flag for detecting the error position by TA). T
A is a phenomenon observed when a magnetoresistive (MR) element is used for the read head, and when the MR element contacts a protrusion on the magnetic disk and the temperature of the element rises due to friction. Since the resistance changes, the abnormal amplitude due to the resistance change due to the heat is superimposed on the original signal component.

As shown in FIG. 3, R through the R / W amplifier
Amplitude of analog read waveform 301 from the head input to the / W channel can be monitored, and when a signal amplitude exceeding a certain slice 302 is detected, it can be estimated that TA exists at that position and an error has occurred there. Is.

The error flag is a flag issued at a portion where an error may occur as its name suggests, and is expected to be R / when the NRZ data decoded from the analog read waveform is correct. The ideal signal to be input to the W channel data recovery circuit and the actual R /
The error of the signal input to the W channel is monitored, and if this error is large, a flag indicating that an error may have occurred at that position, issued by the R / W channel.

Not only the erasure correction using the flag issued by the R / W channel, but also, for example, Japanese Unexamined Patent Publication No.
As disclosed in Japanese Patent Application Laid-Open No. 20403, a method of multiplexing error correction codes and allowing an error in the error correction code by a majority decision circuit, and an error occurrence position virtually as shown in Japanese Patent Laid-Open No. 2000-100806. A method of setting and changing the position to perform erasure correction is also considered.

After obtaining the error position by TA or the like, a corrected error position polynomial is then generated from the syndrome and the disappearance position polynomial. Then, an error value polynomial ω (x) is generated from the corrected error position polynomial. Finally, the route search and error generator 205 performs a route search on the error locator polynomial to detect the error position, calculates the error value polynomial and the modified error locator polynomial, and obtains the error value.

Since the modified error locator polynomial has less unknown constants for specifying the error position than the above-mentioned error locator polynomial, the probability of error correction is high when the same number of errors occur. Therefore, when erasure correction is applied, when data is written with a (2 × n) byte Reed-Solomon code added, a maximum of (2 × n)
It is possible to correct errors up to bytes.

[0014]

When the error position is estimated using the TA flag, it is necessary to set the slice level appropriately in order to improve the accuracy. TA
It is better to set the slice level to a lower level in order to reliably detect its presence at the position where is present, but if it is set too low, even if there is no TA when noise is superimposed on the read signal, etc. There is a possibility that the TA is erroneously detected, that is, the TA flag springs up.
Since the size of TA also changes depending on the shape of the protrusion on the disk, it is a difficult task to set the slice to an appropriate level that does not leak and does not cause erroneous detection.

Further, since error propagation occurs when decoding digital data from the analog read waveform, T
Even if the position of A can be detected correctly, it is difficult to correctly estimate which range is actually in error in the digital data after decoding by the TA.

The error flag is more effective than the TA flag for estimating the error position because it can detect not only TA but also a magnetic recording abnormality. But TA
Similar to the flag, if the slice level of the equalization error that causes the flag to be issued is not set appropriately, there is a problem that the error detection may be skipped or erroneous detection may occur.

Further, in the case of performing the erasure correction by changing the virtual error occurrence position in a detailed manner like the technique described in Japanese Patent Laid-Open No. 2000-100806, in the method of erasure correction, errors of various byte numbers are distributed to a plurality of positions. If it exists, the probability that the virtual error position will hit the actual error position is considerably low.

The present invention provides means for estimating the error position by using the NRZ data itself in order to avoid leakage or spring of the TA flag and the error flag caused by the difficulty of optimally setting the slice as described above. .

[0019]

As a factor causing an error when reproducing NRZ data from an analog read waveform, in addition to TA described above, S /
Examples include N deterioration, decrease in read signal amplitude due to defects on the disk, offset during data writing, fluctuations in waveform amplitude and asymmetry, and changes in read signal waveform characteristics due to changes in operating environment. In the case of an error caused by any factor, if the magnetic disk portion in which the error has occurred is repeatedly read and the obtained analog signal is reproduced as NRZ data, the same position may be mistakenly reproduced every time. Is extremely low. That is, it can be said that the portion that is reproduced in the same way every time is likely to be correct data.

Therefore, a plurality of NRZ data obtained by reading the data recorded on the magnetic disk a plurality of times are stored, and these data are compared in byte units, and the same data is obtained every time or a plurality of times. It is judged that the portion reproduced as is correct data, and conversely, the portion reproduced as different data every time or a plurality of times is a portion where an error has occurred, and that position is used as an erasure correction erasure pointer.

[0021]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the flow of data writing and reading in a general storage device (magnetic disk device) using a magnetic disk as a data recording medium. The flow of data write is as follows.

First, the host sends user data to the magnetic disk device. The magnetic disk device that has received the user data has a hard disk controller (HDC) having a signal processing circuit.
ller) 101, an ECC code for the purpose of error detection and correction and a CRC code for confirmation of correction are generated based on the data. Then, the HDC 101 transfers the NRZ data to which these codes are added to the R / W channel 102. The R / W channel 102 subjects the NRZ data transferred from the HDC 101 to processing such as encoding, scrambling, precoding, and write precompetition to convert the NRZ data into write data suitable for magnetic recording, and the R / W amplifier 103. Send to. The R / W amplifier 103 generates a write current according to the write data. Light head 10
4 records data as a magnetization reversal pattern on the magnetic disk 105 in response to the supplied write current.

When reading the data recorded on the magnetic disk 105, the procedure is as follows. Read head 10
The analog read waveform read at 6 is first amplified by the R / W amplifier 103. Then, the amplified analog signal is sent to the R / W channel 102. R / W channel 1
02 is NRZ which is digital data from analog signal
The data is reproduced and transferred to the HDC 101. HDC10
1 reproduces user data from NRZ data. NR
When reproducing the user data from Z, the ECC code added at the time of writing is used to determine whether the reproduced user data is correct, and if there is an error, the error position is detected and the error is corrected. . Further, the CRC code is used to determine whether the corrected user data is correct, and if it is determined to be correct, the user data is transferred to the host.

The user data transferred from the host to the HDC 101 at the time of writing, or the HDC 10 at the time of reading
The user data reproduced by No. 1 is normally stored in the buffer memory 107 and then transferred to the R / W channel 102 or the host through the HDC 101. In addition, an inductive element is attached to the write head 104 and the read head 10
6 is a recent tendency to use a composite head integrated by using MR (including GMR and TMR) elements.

A Reed-Solomon code is generally used as the ECC code. The error detection and correction procedure using the Reed-Solomon code is as follows. First, the syndrome is generated from the read data. An error-locator polynomial is generated from this syndrome. Next, from the error locator polynomial to the error value polynomial (erro
r-magnitude polynomial) is generated. Then, a root search (chien search) is performed on the error position polynomial to detect the error position. Finally, the error value polynomial is evaluated in the error locator polynomial to calculate the error value. When data is written with a (2 × n) -byte Reed-Solomon code added, it is possible to identify and correct the position of an error of up to n bytes occurring in the read data.

FIG. 4 shows the configuration of the error detection / correction system of the present invention. The part surrounded by the broken line in FIG. 4 is provided as the function of the HDC. The read head 106 repeatedly reads a certain sector on the magnetic disk a plurality of times and outputs a read analog waveform. The R / W channel 401 reproduces the input waveform into a plurality of NRZ data. Then, the vanishing point detection data memory 402 stores the NRZ data. The buffer memory 403 stores one or a plurality of NRZ data among them.

The comparator 404 retrieves a plurality of NRZ data from the vanishing point detection data memory 402 and compares them for each byte. The comparator 404 is shown in FIG.
In the decoder 405, an erasure pointer is issued because an error has occurred at the byte position of the condition described later.
Give to. Alternatively, an erasure pointer may be issued and given to the decoder 405, assuming that an error has occurred at byte positions where the values of a plurality of NRZ data are all different. If they are different, the error may occur at the byte position.

The syndrome generator 406 stores one or a plurality of NRZs stored in the buffer memory 403.
Generate a syndrome from any one of the data,
It is given to the decoder 405. The decoder 405 has a comparator 404.
The erasure position polynomial (eras
ure-locator polynomial), and a modified error locator polynomial and error value polynomial from the syndrome and vanishing locator polynomial. Then, an error value polynomial ω (x) is generated from the corrected error position polynomial. Finally, the route search and error position generator 407 detects an error position by performing a successive substitution search on the corrected error position polynomial, and evaluates the error value polynomial in the corrected error position polynomial to calculate an error value. By applying this modified error locator polynomial, when data is written with a Reed-Solomon code of (2 × n) bytes added, as in the case of using erasure correction, up to (2 × n) bytes can be written. It is possible to correct the error.

Since the error position specifying ability is superior to that of the conventional method, the conventional method has no choice but to retry, or the number of errors is such that the data cannot be read even if the retry is performed. Even if it is, the correct user data can be returned to the host by specifying the error position and combining it with the erasure correction.

Further, as shown in FIG. 5, the buffer memory and the vanishing point detection data memory may be shared. In this case, all of the plurality of NRZ data reproduced on the R / W channel are used as the vanishing point detection data memory / buffer memory 5
Save it in 02. Note that the function of comparing data has a dedicated hardware comparator 40 as shown in FIGS.
4, 503 are not necessarily provided, and software comparison may be performed using a microcomputer.

Since the error correction of the present invention requires reading the data a plurality of times, it takes longer time than the error detection and correction using the ordinary Reed-Solomon code. Therefore, when applying the present invention to a product, it is conceivable to first perform error detection and correction or erasure correction using a general Reed-Solomon code, and then execute it when error detection and correction cannot be performed. .

6 and 7 show a specific flow of the lost pointer detection processing performed in the comparators shown in FIGS. 4 and 5. Here, the procedure for obtaining the vanishing point from the NRZ data obtained by reading a certain sector three times is shown as an example. Note that j in FIGS. 6 and 7 is the number of data bytes in one sector. The number of times the same sector is read to determine the position where an error has occurred is 3 here, but this number is not particularly limited. Depending on the method of issuing the erasure pointer, if it is at least twice, it is possible to specify the error occurrence point, and if the read number is increased, it becomes easier to specify. Therefore, it is possible to change the read count so that the user can set the read count from the outside, and if the lost pointer cannot be issued at the initially set read count, the read is further performed. It is desirable to leave it.

FIG. 6 shows a process in the case of issuing an erasure pointer assuming that an error has occurred at a byte position other than the same value in all of the plurality of NRZ data.

First, k = 0 is set, and detection is performed in order from the first byte of the sector. In step 601, the NRZ data NRZ_A obtained in the first read is compared with the NRZ data NRZ_B obtained in the second read.
If the comparison result is different, it is determined that this byte has an error, and the process proceeds to the next byte.

If the comparison results are the same, in step 602, the NRZ data NRZ obtained in the first read
_A and NRZ data NRZ_ obtained by the third read
Compare with C. If the comparison result is different, it is determined that this byte has an error, and the process proceeds to the next byte. If the result of the comparison also matches in step 602, it is determined that an error has not occurred in the relevant byte,
Increment k by 1 to advance to the next byte.

By executing this until k becomes larger than j, that is, at all bytes, it is possible to know at which byte position the error has occurred, and the lost pointer is obtained.

In this method, the number of errors itself may increase as compared with the case of reading the data once in order to judge that an error occurs if a different value is read even once even among the data compared a plurality of times. However, even if the number of errors increases, the location of the error can be specified, so that the error can be corrected. In other words, when using a Reed-Solomon code of (2 × n) bytes, conventionally, only n bytes could be corrected because n bytes were used to specify the error location, but in the present invention, the location can be specified. , N bytes or more (n + bytes whose location can be specified) can be corrected.

FIG. 7 shows a process in the case of issuing an erasure pointer assuming that an error has occurred at byte positions where the values of a plurality of NRZ data are all different.

In this processing method as well, k = 0 first
Then, the sectors are detected in order from the first byte of the sector. In step 701, the NRZ data NRZ_A obtained in the first read and the NRZ data NRZ_B obtained in the second read are combined with the NRZ data NRZ_A obtained in the first read and the third read in step 702. And the NRZ data NRZ_C obtained in step 7 and step 7
NRZ data NRZ obtained by the second read at 03
_B and NRZ data NRZ_ obtained by the third read
Compare with C.

If the NRZ data do not match in any of these steps, it is determined that an error has occurred in the relevant byte. It is the same as in FIG. 6 until k is greater than j, ie, all bytes are executed.

In the case of this method, unlike the embodiment shown in FIG. 6, if the same value appears in any of the reads, it is determined that the value is correct. Therefore, even if the number of errors exceeds the correction capability in one read, if the number of reads is increased, the number of errors such as misreads will decrease. Therefore, correct the errors after multiple reads. Can be done.

Due to such characteristics, the embodiment of FIG. 7 is effective not only when used alone but also when combined with the embodiment of FIG. First, even if the error pointer is specified in the embodiment of FIG. 6 and the number of errors still exceeds the correction capability and the error cannot be corrected, the embodiment of FIG. 7 may be performed. In addition, when the number of reads is increased, if the data value is the same as the eclectic method of the method of FIG. 6 and the method of FIG. 7 not a total number of times but a certain number (a majority, almost all, etc.) of times. , There is also a method to judge that no error has occurred at that position. In other words, it is a method of determining that an error has occurred at the byte position when the value is different in any of a plurality of digital data. For example, if the same value is obtained eight times out of the number of times of reading 10 times, it is determined that the value is correct, and the error is determined to be another position.

Here, the number of bytes of the added ECC is (2 × N) bytes, and the number of bytes that specifies the error pointer is (2 × M) bytes (where M ≦ N). If the number is L bytes, the number of bytes that can be corrected is (N + ML) bytes.

Since it is impossible to obtain the number of bytes of L in advance, in the present invention, it is assumed that L = 0 and reading is performed a predetermined number of times, and (N + M) ≦ (2 × N), that is, M ≦ N. After correction is performed at the point of time, an error check is performed using the correction confirmation code CRC added to the data together with the ECC to determine whether the corrected data is correct. If the data is incorrect as a result of confirmation by this CRC, (N + M-
If L) ≧ (2 × N), the reading is further performed a predetermined number of times, M ′ (> M) error pointers are obtained, and error correction and CRC confirmation are performed again. By repeating this routine, a maximum of (2 × N) error corrections can be performed in the present invention.

When the conventional error correction method is used for comparison, the error correction method that does not specify the error pointer has M = L = 0 and has only N bytes of correction capability. When an error pointer is virtually set and corrected, the number of bytes that can be corrected is (N +
(M−L) bytes, but L is larger than the present invention. A processing time until (N + ML) ≧ (2 × N) is required as compared with the present invention.

Next, a method for confirming the effect of the correction according to the present invention will be described with reference to FIG. The user data given from the host will now be referred to as user data A801, and the NRZ data to which the Reed Solomon code is added by the HDC 802 when the user data A is given and which is sent to the R / W channel will be referred to as NRZ data A803. The NRZ data A 803 includes a (2 × n) byte Reed-Solomon code 805. In FIG. 8, the Reed-Solomon code is represented as an ECC code.

Next, based on the NRZ data A803, the NRZ
NRZ data B806 and NRZ data C808 having data values different from data A803 by m bytes 804, 807 and 809 are created. Here, m is n <m ≦ (2 × n)
NRZ data B806 and NRZ data C80
8, the m-byte portions 804, 807, and 809 different from the original NRZ data A 803 have the same byte position counted from the data head, and the NRZ data B 806 and the NRZ data C 808 also have m-byte portions different from each other. Must be data. Note that the m-byte portions 803, 807, and 809 are shown as one location in FIG. 8, but may be divided into a plurality of portions as long as the total number of bytes is m bytes.

Next, the write data when the NRZ data B806 and the NRZ data C808 are input to the R / W channel 810 will be referred to as write data B811 and write data C812, respectively. And these write data B8
11 and write data C 812 are applied to the R / W amplifier 813, respectively.
814 and write current waveform C815. Further, the analog read waveforms when the magnetization reversal, which occurs when data is written on the magnetic disk with these write current waveforms, are read by the read head are referred to as analog read waveform B817 and analog read waveform C818, respectively.

In each process of generating write data from NRZ data, write current waveform from write data, magnetization reversal from write current waveform, and analog read waveform from magnetization reversal, bit omission, bit spring, timing jitter, and amplitude abnormality are included. Generate write data and analog read waveforms so that they do not exist.

When the analog read waveform B817 and the analog read waveform C818 are given to the read circuit which only performs the correction by the ordinary Reed-Solomon code, both of them contain an error of more than n bytes, and therefore they are repeated or continuously. Even if these analog read waveforms are given, correct user data cannot be reproduced. Further, since the analog read waveform B817 and the analog read waveform C818 created by changing the NRZ data do not include TA or magnetic recording abnormality, no TA flag or error flag is issued.
Therefore, correct user data cannot be reproduced even if these input waveforms are given to the reproducing system which is performing the erasure correction using the flag issued by the R / W channel.

However, in the case of a reproducing system performing erasure correction using the present invention, the NRZ data B and NRZ data C reproduced from the analog read waveform B and the analog read waveform C by R / W differ in m bytes. And that part is the part where the error has occurred. Therefore, by providing this information to the decoder as an erasure pointer, erasure correction can be performed and correct user data, that is, user data A801 can be reproduced.

Note that here, the two simplest examples, two N
Although the case where the vanishing point is obtained from the RZ data is shown,
Even when the vanishing point is obtained from three or more NRZ data, the NRZ data A801 to the NRZ data B806
The operation of erasure correction can be confirmed by creating a plurality of NRZ data by the same method as that used to create the NRZ data C808 and inputting them into the reproducing system.

As a result of applying the present invention, T which is generated when the TA flag or the error flag is used,
It becomes possible to eliminate erroneous detection of A or error and omission of detection, and to correctly specify the error occurrence position, so that it becomes possible to perform erasure correction more effectively and it seems that it could not be read in the past. Even data can be read correctly.

[Brief description of drawings]

FIG. 1 is a diagram showing a general configuration of a magnetic disk device.

FIG. 2 is a diagram showing a configuration of a conventional magnetic disk device having an erasure correction function.

FIG. 3 is a diagram showing how TA is detected from an analog read waveform.

FIG. 4 is a diagram showing an error correction system configuration in a case where a vanishing point detection data memory is provided separately from a buffer memory, which is one embodiment of the present invention.

FIG. 5 is a diagram showing an error correction system configuration when the vanishing point detection data memory is also used as a buffer memory.

FIG. 6 is a flowchart showing a procedure for determining that an error has occurred at the byte position if even one of a plurality of NRZ data has a different value.

FIG. 7 shows the case where the values of a plurality of NRZ data are all different,
It is a flow chart which shows a procedure which judges that an error has occurred at the byte position.

FIG. 8 is a diagram for explaining a procedure for creating confirmation data.

[Explanation of symbols]

101 ... HDC, 102 ... R / W channel, 103 ... R
/ W amplifier, 104 ... Write head, 105 ... Magnetic disk, 106 ... Read head, 107 ... Buffer memory, 201 ... R / W channel, 202 ... Buffer memory, 203 ... Syndrome generator, 204 ... Decoder, 2
05 ... Route search and error value generator, 206 ... HD
C, 301 ... Analog read waveform, 302 ... TA detection slice level, 401 ... R / W channel, 402 ... Vanishing point detection data memory, 403 ... Buffer memory, 4
04 ... Comparator, 405 ... Decoder, 406 ... Syndrome generator, 407 ... Route search and error value generator, 4
08 ... HDC, 501 ... R / W channel, 502 ... Vanishing point detection data memory / buffer memory, 503 ... Comparator, 504 ... Decoder, 505 ... Syndrome generator, 5
06 ... Route search and error value generator, 507 ... HD
C, 801 ... User data A, 802 ... HDC, 803
... NRZ data A, 804 ... m-byte portion, 805 ...
Error correction code, 806 ... NRZ data B, 807 ... m
Byte portion, 808 ... NRZ data C, 809 ... m byte portion, 810 ... R / W channel, 811 ... Write data B, 812 ... Write data C, 813 ... R / W amplifier, 814 ... Write current waveform B, 815 ... Write current waveform C, 816 ... Head / magnetic disk, 817 ... Analog read waveform B, 818 ... Analog read waveform C.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G11B 20/18 G11B 20/18 572F 576 576Z G06F 11/10 330 G06F 11/10 330P G11B 20/10 321 G11B 20/10 321Z H03M 13/15 H03M 13/15

Claims (8)

[Claims] 1. A recording medium, a head for reading data recorded on the recording medium, an amplifier for amplifying a signal read by the head, and digital data reproduced from the read signal amplified by the amplifier. A storage device comprising a channel and a comparator for comparing a plurality of digital data values at the same byte position obtained by reading from the head a plurality of times. 2. A recording medium, a head for reading data recorded on the recording medium, an amplifier for amplifying a signal read by the head, and digital data reproduced from the read signal amplified by the amplifier. An error position is determined by comparing a channel, a data memory that stores a plurality of digital data read by the head a plurality of times and stored in the channel, and a value at the same byte position of the plurality of digital data retrieved from the memory. A erasure pointer, a buffer memory for storing at least one digital data reproduced in the channel, a syndrome generator for generating a syndrome from the digital data stored in the buffer memory, and the erasure pointer. Generate an erasure position polynomial from Memory device comprising a decoder for generating a corrected error position polynomial and error value polynomial from the formula and the syndromes, and error position generator for calculating an error from said corrected error position polynomial and the error value polynomial. 3. An error correction method for a storage device, wherein the same sector is read a plurality of times when the data recorded on the storage medium is reproduced, and a plurality of the obtained data are stored, and the plurality of stored data are stored. An error correction method for a storage device that compares data values of bytes located at the same place in data, determines that an error has occurred in the data at a position where this value is different, and corrects the error based on the determined position. 4. The sector is read three times or more, the values of the data located at the same location are compared each time, and it is determined that an error has occurred in the data at a position where this value is different each time. 4. The error correction method for a storage device according to claim 3, wherein the error is corrected based on 5. The sector is read three times or more, the values of the data located at the same location are compared each time, and it is determined that an error has occurred in the data at a position where this value is different even once, and this determination is made. 4. The error correction method for a storage device according to claim 3, wherein the error is corrected based on the determined position. 6. When reproducing data, the same sector is read out a plurality of times, a plurality of digital data converted from the obtained signal are stored, and the same location from the beginning of the plurality of stored digital data is stored. An error correction method for a storage device that compares the values of the data located at a position for each byte and judges that an error has occurred in the data at the byte position where the values differ, and stores the error information. 7. A method of reading data recorded on a recording medium, amplifying the read signal, reproducing digital data from the amplified signal, and storing a plurality of digital data read a plurality of times. The error position is determined by comparing the values of the same byte positions of the digital data, the at least one reproduced digital data is stored, the syndrome is generated from the stored digital data, and the disappearance position is generated from the disappearance pointer. An error correction method for a storage device, wherein a polynomial is generated, a corrected error position polynomial and an error value polynomial are generated from the vanishing position polynomial and the syndrome, and an error is calculated from the corrected error position polynomial and the error value polynomial. 8. The error correction method for a storage device according to claim 3, wherein an error is corrected using erasure correction based on the determined position.